Hey, I’m zesen 👋
I make AI explainers
and break down how things actually work
Mostly AI, sometimes not
If I can’t source it, I don’t say it
Looking to connect with people
who care about the details, not just the hype
If you’re into AI or tech deep-dives
happy to mutual — just say hi 🙌
Sam Altman just said there’s “no room for debate” that humans have to stay in control of AI.
He’s been saying versions of this a lot lately.
At the same time the labs keep shipping faster models and talking about racing ahead.
It’s interesting watching the same people say “slow down” and “go faster” within a few weeks of each other.
Do you think they actually mean the slowdown part, or is it mostly for the cameras?
Anthropic just open-sourced a prototype called FlashProve.
It can prove a reply actually came from a specific model without showing the weights.
Basically a receipt that says “yes, this was really Claude (or whatever) and not some cheaper thing swapped in.”
Still early and kinda heavy to run, but the idea of being able to check that feels overdue.
Would you actually care about that kind of proof, or is it overthinking things?
@rodesia The mechanism generalises to any hidden UI fetching at full resolution. Fix it at the data layer: store a 400px variant on upload so the small file always exists and nobody can regress by pointing at the original.
@_Dev_Arora__ All four are defensible: Objectivity, God, Perfection and History are each a one word thesis with a whole philosophy behind it. There is no ground truth here, so the variance is the answer.